SPECIAL ARTICLEEconomic & Political Weekly EPW april 26, 2008107Do Interest Rates Matter for Stock Markets? Chakradhara PandaChakradhara Panda (cpanda@mail.tapmi.org) is at the TA Pai Management Institute at Manipal.This paper asks whether interest rates matter for stock markets in the Indian context. It uses the monthly averages of the SENSEXandNIFTY to measure stock prices in April 1996-June 2006. For the same period, the month-end yields on 10-year government security and treasury bills (15-91 days) are used to measure long-term and short-term interest rates, respectively. The paper finds that there is a long-run relationship between interest rates and stock prices. Both long-term and short-term interest rates affect stock prices. The long-term interest rates are found to affect stock prices negatively, whereas short-term interest rates affect stock prices positively. In addition, the SENSEX is found tobe more responsive to changes in interest rates thantheNIFTY.The stock market has become an important indicator of the performance of the Indian economy over the years. With this, the working of the stock market has become a vital and facilitating subject for academics, investment professionals, and monetary policymakers. The stock market works with the sentiments of participants, which depend on several factors, making it a very sensitive segment of the economy. Globalisation and financial sector reforms have added to the sensitivity by increasing determinants of the stock market movement manifold. Among the several other important determinants, one whose relationship with the stock market is of great concern and needs to be carefully studied is interest rates. In theory, the relationship between interest rates and stock prices is negative. This is due to the cash flow discounting model according to which, present values of stocks are calculated by discounting the future cash flows at a discount rate. If the discount rate increases, then present values of stocks decline and vice versa. This discount rate is a risk adjusted required rate of return and equal to the level of interest rates in the economy. Therefore, an increase in interest rates lowers present values of stocks directly. Even a relatively small rise in interest rates can have a major effect on present values if it is spread out over several years. In addition, rising interest rates reduce cash flows by reducing the profitability of firms. Due to these two reasons, present values of stocks decline and so do current stock prices. The inverse holds true as well. Apart from the above theoretical reason, there are other few reasons, which account for the negative relationship between interest rates and stock prices. First, interest rates are risk free returns on bonds and as interest rates on bonds rise, bonds become more attractive and stocks less attractive. Consequently, there is a change in the asset allocation in favour of bonds rather than stocks. This moves funds from the stock market to the bond market, which invariably increases the demand for bonds and reduces the demand for stocks. As a result, the prices of stocks fall. The opposite is true when interest rates fall and funds are shifted from the bond market to the stock market. Second, corporate profitability is hit because of increase in interest rates in two ways: (i) companies earnings net of interest rates fall; and (ii) consumers’ demand for the products decreases they pay more to borrow money. As the profitability decreases stock prices decline and vice versa. Third, if interest rates increase then investors’ expectations about the economy and company earnings, which drive the stock market turn negative. This pushes stock prices down and the reverse is true as well. The effects of interest rates changes on a stock’s intrinsic value are more complex than outlined earlier because of the existence 
GSEC 10 SENSEX 
8 6 4 2 0 SENSEX TB 15 91 
GSEC 5 SENSEX 
GSEC 10 NIFTY 
TB 15 91 NIFTY 
16 14 12 10 8 6 4 2 0 GSEC 5 NIFTY SPECIAL ARTICLE
Figure 3: Movements in Monthly Turnovers of BSE, Government of India Dated Securities, Treasury Bill 91 Day, and Call Money
(Turnover in Rs ‘000 crore) 600
500 GOIDS
400
300
200
100
BSE
0 Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr 1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006
120
100
80
60
40 BSE
20
TB-91
0 Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr 1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006
120
100
80 BSE
60
40 CALL
20
0
Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr
1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006 Sources: Stock Exchange, Mumbai (BSE), Reserve Bank of India, Handbook of Statistics on Indian Economyand RBI Bulletin (various issues).
2 Interest Rates and Stock Markets in India
In this section, a preliminary analysis of the relationship between interest rates and stock markets in India is made by using various facts and figures collected from the Bombay Stock Exchange (BSE), National Stock Exchange of India (NSE), and Reserve Bank of India (RBI). Figures 1 and 2 (pp 108, 109) show movements in stock prices and various interest rates during May 1996-June 2006. The comovements of the BSE sensitive index (SENSEX) with yields on 10-year government security (GSEC-10), 5-year government security (GSEC-5) and 15-91 days treasury bill (TB 15-91) are shown in Figure 1. In the figure, the SENSEX is showing a rising trend over the period. On the other hand, yields on GSEC-10, GSEC-5, and TB 15-91 are generally exhibiting falling trends for the same period barring the period April 2004-June 2006 when the yields firmed up. The declining trend of yields on government securities is primarily contributed to ample liquidity and
110 expectations of interest rates cuts over the period. The firming up of yields from April 2004 is because of the upturn in the international interest rate cycle, rise in international crude oil prices, domestic monetary policy tightening and edging up of inflation. The SENSEX rose from a low level of 3,732 points in May 1996 to reach a peak level of 11,741 points in April 2006. On the other hand, yields on GSEC-10, GSEC-5 and TB 15-91 fell from higher levels of 13.93 per cent, 13.66 per cent, and 11.75 per cent, respectively to lower levels of 7.39 per cent, 6.96 per cent and 5.51 per cent, respectively for the corresponding period. From this it may be concluded that the comovement between SENSEX and yields on GSEC-10, GSEC-5, and TB 15-91 is negative. The negative comovement between SENSEX and yields on various government securities can be gauged from their negative correlation coefficients which are given in Table 1 (p 108). These cor relation
Figure 4: Movements in Monthly Turnovers of NSE, Government of India Dated Securities, Treasury Bill 91 Day, and Call Money
(Turnover in Rs ‘000 crore)
600
500
400
GOIDS
300
200
100
NSE
0 Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr 1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006
250 NSE
200
150
100
50 TB-91
0 Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr 1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006
250 NSE
200
150
100
50 CALL
0 Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr Oct Apr 1998 1998 1999 1999 2000 2000 2001 2001 2002 2002 2003 2003 2004 2004 2005 2005 2006
Sources: National Stock Exchange of India Limited, Mumbai (NSE) and Handbook of Statistics on Indian Economy and RBI Bulletin, Reserve Bank of India (various issues).
april 26, 2008 EPW Economic & Political Weekly
coefficients are calculated over the period May 1996-June 2006. The correlation coefficient bet ween SENSEX and GSEC-10 is found to be -0.4075. Similarly, cor relation coefficients of SENSEX with GSEC-5 and TB 15-91 are -0.4115 and -0.3232 respectively.
The comovements of NSE S&P CNX Nifty (NIFTY) with yields on GSEC-10, GSEC-5 and TB 15-91 are shown in Figure 2. We observe a
negative comovement and Yield on 10-Year Government Security of NIFTY with yields on Figure 5: Monthly Movements in BSE Turnover
(Turnover in Rs ‘000 crore)
SPECIAL ARTICLE
-0.2430. Similarly, turnovers of TB-91 and call are found to be negatively correlated with turnover of BSE for which correlation coefficients are equal to -0.1269 and -0.0593, respectively.
Figure 4 displays comovements of NSE turnover with turnovers in GOIDS, TB-91 and call during April 1998-June 2006. The figure reveals that there is a negative comovement between the turnover of NSE and turnover in GOIDS. For example, turnover of NSE fell from Rs 1,25,347 crore to Rs 85,346 crore during August
various government 2000-August 2003. For the same period, however, turnover in
5 25 45 65 85 105 125 April 1998
securities. The correla-GOIDS increased from Rs 38,347 crore to Rs 4,98,818 crore. For tion coefficients of the rest of the period, turnover of NSE exhibited a rising trend in
Oct 1998
BSE Turnover NIFTY with yields on contrast to a falling trend observed in turnover of GOIDS.
GSEC-10, GSEC-5 and TB Similarly, turnover in call showed a general declining trend 15-91 are negative, opposed to a rising trend in turnover of NSE for the entire period.
April 1999
Oct 1999
which are equal to The negative comovements are further justified on the basis of -0.4847, -0.4773 and their negative correlations given in Table 1. The only exception is
April 2000
-0.3681, respectively the comovement between turnover in NSE and turnover in TB-91
| Oct 2000 | (Table | 1). | It | is | also | ||
|---|---|---|---|---|---|---|---|
| observed in Table 1 that | |||||||
| April 2001 | the | NIFTY | is | more | |||
| Oct 2001 | GSEC-10 | negatively correlated | |||||
| with yields on GSEC-10, | |||||||
| April 2002 | GSEC-5 | and TB | 15-91 | ||||
| than the SENSEX. | |||||||
| Oct 2002 | Figures 3 and 4 | ||||||
| April 2003 | (p 110) | show | move | ||||
| ments in monthly turn- | |||||||
| Oct 2003 | overs in stock markets, | ||||||
| April 2004 | government securities and call money mar- | ||||||
| Oct 2004 | ket. The comovements | ||||||
| of turn over of the BSE | |||||||
| April 2005 | with government of | ||||||
| Oct 2005 | India dated securities | ||||||
| (GOIDS), treasury bill | |||||||
| April 2006 | 91 day (TB-91) and call | ||||||
| 2 | 4 6 8 Yields (%) | 10 12 14 | money (CALL) for the period April 1998-June |
Sources: Stock Exchange, Mumbai (BSE), Reserve Bank of India, Handbook of Statistics on Indian Economy and RBI Bulletin
2006 are shown in
(various issues).
Figure 3. In the figure, the comovement between turnovers of BSE and GOIDS is seen to be negative. For example, turnover of GOIDS, amidst volatility, rose to a high level of Rs 4,98,818 crore in August 2003 from a low level of Rs 38,347 crore in August 2000. However, during the same period, turnover of BSE fell from a higher level of Rs 92,562 crore to reach Rs 36,334 crore. During the rest of the period, turnover of GOIDS showed a declining trend while BSE turnover exhibited a rising trend. However, turnover of BSE in February 2006 was reported to be the lowest in the whole period which was equal to Rs 7,070 crore.
Similarly, Figure 3 exhibits negative comovements between turnover of BSE and turnovers of TB-91 and call during April 1998-June 2006. The negative comovements of BSE turnover with turnovers of GOIDS, TB-91, and call can also be seen from their correlations given in Table 1. The correlation between turnover of BSE and turnover of GOIDS is found to be negative i e,
Economic & Political Weekly EPW april 26, 2008
which is found to be positive (Figure 4). The correlation co efficient between NSE turnover and turnover in TB-91 is positive, which is equal to 0.1442 (Table 1).
The comovements of the turnover of BSE with yields on GSEC-10 and TB 15-91 are respectively shown in Figures 5 and 6 for the period April 1998-June 2006. It is observed in both figures that the turnover of BSE varies Figure 6: Monthly Movements in BSE Turnover and
Yield on 15-91 Days Treasury Bill
positively with yields on
(Turnover in Rs ‘000 crore)GSEC-10 and TB 15-91. 5 25 45 65 85 105 125
April 1998
In other words, the BSE
BSE Turnover
seems to be active and
Oct 1998
liquid when the rates are rising but turn lack-April 1999 lustre and illiquid when
TB 15-91 Oct 1999
the rates fall. The estimated cor relation coef-April 2000 ficient between the turn-
Oct 2000
over of BSE and yield on GSEC-10 is found to
April 2001
be 0.11, whereas it is
0.22 for the correlation Oct 2001 bet ween BSE turnover
April 2002
and yield on TB 15-91. On the other hand, the
Oct 2002
turn over of NSE is found
April 2003
to move in versely in general with yields on
Oct 2003
GSEC-10 and TB 15-91, which are shown in April 2004 Figures 7 and 8 (p 112),
Oct 2004
respectively. The correlation co efficients of
April 2005
turnover of NSE with yields on GSEC-10 and Oct 2005 TB 15-91 are found to be
April 2006
negative during April
2 4 6 10
8 12 1998-June 2006, which Yields (%)
Sources: Stock Exchange, Mumbai (BSE), Reserve Bank of India,
are equal to -0.29 and
Handbook of Statistics on Indian Economy and RBI Bulletin -0.19, respectively. (various issues).
111
SPECIAL ARTICLE
The monthly move ments in interest rates and stock markets price-earning (P/E) ratios are shown in Figures 9 to 12 (p 113). Figure 9 displays the negative comovement between the P/E ratio for the 30 scrips included in the BSESENSEX and yield on GSEC-10
during May 1996-
Figure 7: Monthly Movements in NSE Turnover and Yield on 10-Year Government Security
June 2006. Whenever
(Turnover in Rs ‘000 crore)
yield on GSEC-10 has
10 45 80 115 150 185 220 April 1998
decreased, the BSE SENSEX has a higher P/E
Oct 1998
ratio. This is reflected in the negative, although
April 1999 NSE Turnover
low, correlation coeffi-
Oct 1999
cient between the two
GSEC-10
which is equal to -0.01.
April 2000
However, on the other hand, yield on TB 15-91 shows a positive cor-
Oct 2000
April 2001
relation with the BSE SENSEXP/E ratio during
Oct 2001
the same period, which
April 2002
is shown in Figure 10. The correlation coeffi-
Oct 2002
cient bet ween these two is positive and equal to
April 2003
0.20. Figure 11 shows
Oct 2003
the positive comovement bet ween NIFTY
April 2004
P/E ratio and yield on GSEC-10 during January 1999-April 2006. The
Oct 2004
April 2005
correlation co efficient bet ween these two is
Oct 2005
found to be 0.53. Simi-
April 2006
larly, in Figure 12, we
2 4 6 8 10 12 14
see a positive comove-
Yields (%)
ment bet ween P/E ratio
Sources: National Stock Exchange of India Limited, Mumbai (NSE) and Handbook of Statistics on Indian Economy and RBI Bulletin,
of NIFTY and yield on TB
Reserve Bank of India (various issues).
15-91 during January 1999-April 2006. A positive correlation coefficient, which is equal to 0.59, is observed between NIFTYP/E ratio and TB 15-91 yield.
3 Data and Methodology
Monthly data series for the period from April 1996 to June 2006 are used in this study. The total number of observations, which is equal to 121, is believed to constitute a large data set for any kind of time series analysis. We use the monthly averages of the BSE SENSEX and NIFTY to measure stock prices. The choice is made with the obvious belief that these two indices are the pulse of the Indian stock markets. The month-end yields on GSEC-10 is used to measure long-term interest rates while the month-end yields on TB 15-91 is taken to represent short-term interest rates. Stock prices i e, the SENSEX and NIFTY are expressed in logarithmic forms for the analysis. The same is not done for long-term and short-term interest rates. This approach is standard as transformation of interest rates which are expressed in percentages into logarithms may add complications to the interpretation. Data on
112 stock prices are obtained from the BSE and NSE. The data of 10-year government securities and TB-15-91 are obtained from various issues of Handbook of Statistics on the Indian Economy and RBIBulletin published by the RBI.
The cointegration methodology is employed to investigate the long-run relationship among stock prices, short-term interest rates and long-term interest rates. Before employing the cointegration technique, it is required to pretest the variables for their order of integration. This is because in the cointegration all the variables are required to be integrated of the same order. The augmented Dickey-Fuller (ADF) and Phillips-Perron (PP) unit root tests are used in this paper to infer the order of integration in each of the variables. If all variables are stationary i e, integration of order o, it is not necessary to employ the cointegration methodology since standard time-series methods apply to stationary variables. If the variables are found to be integrated of dif ferent orders then it can possibly be concluded that they are not cointegrated.
In the case all variables are found to be integrated of the same order, say integrated of order 1, we proceed with cointegration test in the Johansen (1988) and Johansen and Juselius (1990) framework.1 This framework allows for the testing of more than one cointegrating vector in the data by calculating the maximum likelihood estimates on these vectors. Two test statistics such as
λtrace and λmax are used Figure 8: Monthly Movements in NSE Turnover and Yield on 15-91 Days Treasury Bill
in order to determine
(Turnover in Rs ‘000 crore)
| the number of cointe | 10 | 45 | 80 115 | 150 | 185 | 220 | |
|---|---|---|---|---|---|---|---|
| grating vectors. | April 1998 | ||||||
| Johansen and Juselius | Oct 1998 | ||||||
| (1990) provide the criti- | NSE Turnover | TB 15-91 |
April 1999
cal values of the λtrace and λmax statistics. If the
Oct 1999
test statistic is greater than the critical value at
April 2000
a significance level then
Oct 2000
the null hypothesis of r cointegrating vectors is April 2001 rejected in favour of the
Oct 2001
alternative hypothesis. If the variables are
April 2002
found to be cointe -
Oct 2002
grated i
e, the longrun relation ship exists
April 2003
among variables, the vector error correc-Oct 2003 tion model (VECM) can
April 2004
be employed to establish the Granger causal Oct 2004 direction. VECM allows
April 2005
the modelling of both the short-run and long-
Oct 2005
run dynamics for the
April 2006
variables involved in the model. Engle and
2 4 6 8 10 12 Yields (%) Sources: National Stock Exchange of India Limited, Mumbai (NSE)
Granger (1987) show
and Handbook of Statistics on Indian Economy and RBI Bulletin,that cointegration is Reserve Bank of India (various issues).
april 26, 2008 EPW Economic & Political Weekly
implied by the existence of a corresponding error correction representation which implies that changes in the dependent variable are a function of the level of the disequilibrium in
Figure 9: Monthly Movements in BSE Sensex Price-Earning Ratio and Yield on 10-Year Government Security
| 35 – | – 16 | ||
| 30 – | GSEC-10 | – 14 | |
| – 12 |
25 –
SPECIAL ARTICLE
disturbances. The lag length i is determined by using the likelihood ratio (LR) test.
4 Empirical Results and Discussion
As a first step, unit root tests are conducted in order to establish the order of integration for stock prices and interest rates. The ADF and PP unit root tests are used for this purpose. Table 2 shows the results of unit root tests for four variables such as SENSEX, NIFTY,
P/E ratio
– 10
20 –
Yield (%) Yield (%)
GSEC-10 and TB 15-91 in levels and first differences. All variables
P/E ratio
– 8
15 – – 6
are found to be non-stationary (i e, presence of unit root) in levels
10 –
– 4
5 – – 2 May Aug Aug Aug Aug Sep Sep Sep Sep Oct 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 Sources: Stock Exchange, Mumbai (BSE), Reserve Bank of India, Handbook of Statistics on Indian Economy and RBI Bulletin (various issues).
| 35 – | – 16 | |||
| 30 – | – 14 | |||
| 25 – | TB 15-91 | P/E ratio | – 12 |
according to the ADF test. However, they are stationary (i e, rejection of presence of unit root) on their first differences. So far as the PP test is concerned, all variables except TB 15-91 are found to contain a unit root in levels. Dua et al (2003) apply three unit root tests such as ADF, PP and Kwiatkowski, Phillips, Schmidt and Suin (KPSS) to test the presence of a unit root in TB 15-91. Except the PP, the other two tests provide the evidence of a unit root in levels of TB 15-91. On the basis of two out of three tests supporting for the
Table 2: Unit Root Test Results
Variables Lags ADF
PP
P/E ratio
– 10
20 –
– 8
15 –
– 6
Levels First Differences Levels First Differences
| 10 – | – 4 | |||||||||
| 5 – May 1996 | Aug 1997 | Aug 1998 | Aug 1999 | Aug 2000 | Sep 2001 | Sep 2002 | Sep 2003 | Sep 2004 | Oct 2005 | – 2 |
Sources: Stock Exchange, Mumbai (BSE), Reserve Bank of India, Handbook of Statistics on Indian Economyand RBI Bulletin (various issues).
the cointegrating relationships (captured by error correction term) and changes in other independent variables. According to Granger representation theorem, if variables are cointegrated then their relationships can be expressed as ECM. Provided that variables in our case are cointegrated of order r, the VECM can be written as:
ΔLSENSEXt = α1+αLSENSEXêt–1 + Σα11(i) ΔLSENSEXt–i + i=1
Σα12 (i) ΔLNIFTYt–i + Σα13 (i) ΔGSEC-10t–i + i=1 i=1
| LSENSEX 1 | -1.2903 | -5.8657** -1.3176 -10.0259** | |
|---|---|---|---|
| 4 | -1.2599 | -4.3257** -1.4977 -10.0798** | |
| LNIFTY | 8 1 | -1.4438 -1.7892 | -3.8790* -1.5616 -10.0807** -6.0172** -1.504 -8.1202** |
| 4 | -1.476 -4.2233** -1.6561 -8.0865** | ||
| 8 GSEC-10 1 | -1.5595 -0.5017 | -3.8298* -1.6674 -7.9987** -7.6474** -0.5442 -10.9535** | |
| 4 8 | -0.7656 -3.8587* -0.5392 -10.9555**-0.624 -3.0509 -0.6242 -10.9524** | ||
| TB 15-91 | 1 | -2.8736 | -9.5174** -4.0486** -15.4599** |
| 4 8 | -2.3038 -2.8114 | -5.6309** -4.2829** -16.5638**-4.3044** -4.7139** -17.5015** |
** and * indicate the rejection of hypothesis of a unit root at 1 per cent and 5 per cent levels respectively. The MacKinnon critical values for rejection of the hypothesis of a unit root for ADF at 1 per cent and 5 per cent levels are -4.0429 and -3.4504, respectively; and for PP at 1 per cent and 5 per cent levels are -4.0400 and -3.4491, respectively. The test regressions for ADF and PP include a constant and linear trend. L indicates the logarithmic form of the variable.
Σα14 (i) ΔTB15-91t–i +e1t ...(1) i=1
Figure 11: Monthly Movements in NSE S&P CNX Nifty Price-Earning Ratio and Yield on 10-Year Government Security
ΔLNiftyt = α2+αLNiftyêt–1 + Σα21(i) ΔLSENSEXt–i + i=1
GSEC-10 – 14 30 – P/E ratio
Σα22 (i) ΔLNIFTYt–i + Σα23 (i) ΔGSEC-10t–i +
– 12
i=1 i=125 –
Yield (%)
– 10
Σα24 (i) ΔTB15-91t–i +e2ti=1
P/E ratio
...(2)
20 –
– 8
15 –
– 6
ΔGSEC-10t = α3+αGSEC-10êt–1 + Σα31(i) ΔLSENSEXt–i +
i=110 –
– 4
Σα32 (i) ΔLNIFTYt–i +Σα33 (i) ΔGSEC-10t–i + 5 – – 2 i=1 i=1Jan Jan Jan Jan Jan Jan Jan Jan 1999 2000 2001 2002 2003 2004 2005 2006
Σα34 (i) ΔTB15-91t–i +e3t ...(3)
Sources: National Stock Exchange of India Limited, Mumbai (NSE) and Handbook of Statistics on Indian Economy and
i=1
RBI Bulletin, Reserve Bank of India (various issues).
ΔTB15-91t = α4 + αTB15-91êt–1 + Σα41(i) ΔLSENSEXt–i +Figure 12: Monthly Movements in NSE S&P CNX Nifty Price-Earning Ratio
i=1
and Yield on 15-91 Days Treasury Bill
Σα42 (i) ΔLNIFTYt–i + Σα43 (i) ΔGSEC-10t–i + i=1 i=1
30 – – 14 TB 15-91 P/E ratio
– 12
Σα44 (i) ΔTB15-91t–i +e4t ...(4)
25 –
i=1
Yield (%)
– 10
20 –
– 8
15 –
P/E ratio
where the error correction term êt–1 represents the previous
period’s deviation from long-run equilibrium. αLSENSEx, αLNIFTY,
– 6
10 –
– 4
αGSEC-10 and αTB 15-91 coefficients are called the speed of adjust
5 – – 2
ment. These coefficients represent the proportion by which
Jan Jan Jan Jan Jan Jan Jan Jan 1999 2000 2001 2002 2003 2004 2005 2006
the long-run disequilibrium in the dependent variables is
Sources: National Stock Exchange of India Limited, Mumbai (NSE) and Handbook of Statistics on Indian Economy and corrected in each short period. e1t, e2t, e3t and e4tare white noise RBI Bulletin, Reserve Bank of India (various issues).
Economic & Political Weekly EPW april 26, 2008 113
SPECIAL ARTICLEapril 26, 2008 EPW Economic & Political Weekly114non-stationarity of TB 15-91, they conclude that TB 15-91 is non-stationary in levels. Following this finding, we consider TB 15-91 to be non-stationary in levels. Meanwhile, the null hypothesis of a unit root is strongly rejected for all variables on first differences according to the PP test. To summarise the results of the ADF and PP tests, it can be said that all variables are found to be integrated of order one i e, I(1). Next, the Johansen cointegration test is applied to check the cointegration between stock prices and interest rates. The number of lags in the model is optimally chosen as seven by the LR test. The results of Johansen’s cointegration test are presented in Table 3. The results indicate that there exists at lest one cointegrating relation-ship betweenSENSEX, NIFTY, GSEC-10 and TB 15-91. This is because the calculated values ofλtrace and λmax test statistics exceed the 5 per cent critical values for testing the null hypothesis of a zero cointe-grating vector. Therefore, the null hypothesis is rejected in favour of the alternative hypothesis of at least one cointegrating vector. The presence of at least one cointegrating vector indicates that there exists a long-run relationship betweenSENSEX, NIFTY, GSEC-10 and TB 15-91. Hence, the vector error correction model can be established. The results of the vector error correction model are shown in Table 4. The results show that the error correction term is significant in equations 1 and 2. In other words, the lagged error correction is statistically significant in the SENSEX andNIFTY equation. However, the error correction term is insignificant in theGSEC-10 and TB 15-91 equation. These results imply that the SENSEX andNIFTY change in response to stochastic shocks (represented by e1tand e2t in equations 1 and 2, respectively) and to the previous period’s deviation from long-run equilibrium. The sizes of error correction coefficients of the SENSEX andNIFTY are equal to -0.6032 and -0.2964, respectively. This finding suggests that theSENSEX andNIFTY fall to correct the long-run disequilibrium. The SENSEX corrects 60.32 per cent andNIFTY corrects 29.64 per cent every month so as to attain the long-run equilibrium. The significant error correction term in the case of the SENSEX and NIFTY and insignificant error correction term in the case of the GSEC-10 and TB 15-91 have another important implication. This result conveys that interest rates such as GSEC-10 and TB 15-91 do exert independent influences on stock prices ie, the SENSEX and NIFTY. In other words, this implies a unidirectional long-run causality from interest rates towards stock prices. In addition to the long-run causality, the short-run causality is also observed from interest rates to stock prices. This is because the five-period lagged difference of GSEC-10 is significant in explaining changes in the SENSEX and NIFTY. The significant negative coefficients of the five-period lagged difference of GSEC-10 with respect to the SENSEX and NIFTY, which are equal to -0.0360 and -0.0277, respectively imply that long-term interest rates affect stock prices negatively in the short-run. However, the change in GSEC-10 is shown in the table to be affected by the six-period lagged difference of the NIFTY. From this, it may be concluded that there is short-run bidirectional causality between GSEC-10 and NIFTY. As far as the short run causality between short-term interest rates and stock prices is concerned, Table 4 shows a unidirectional causality from TB 15-91 towards the SENSEX and NIFTY. For exam-ple, the three-period, four-period and five-period lagged differ-ences of TB 15-91 are found to be significant in explaining changes in the SENSEX. The significant coefficients of TB 15-91 are posi-tive, which, in turn, implies that short-term interest rates affect the SENSEX positively. Similarly, the three-period lagged differ-ence of TB 15-91 is significant and affects the NIFTY positively. From this, it may be surmised that short-term interest rates af-fect stock prices positively in the short-run. The above findings suggest that both long-term and short-term interest rates affect stock prices. The long-term interest rates affect stock prices negatively whereas short-term interest rates are found to affect them positively. Intui-tively a negative effect of interest rates is expected due to the direct effect of on the discount rate. The long-term interest rates directly reflect the discount rate and hence should have a negative Table 4: Parameter Estimates of Vector Error Correction ModelIndependent DependentVariable Variable ΔLSENSEX ΔLNIFTY ΔGSEC-10 ΔTB 15-91Constant -0.0023 [-0.6607] 0.0005 [0.1611] -0.0376 [-0.8448] -0.0777 [-0.6444]êt–1 -0.6032** [-4.4459] -0.2964** [-2.9281] 1.4186 [0.8187] -2.8025 [-0.5968]ΔLSENSEX(-1) 0.0794 [0.4672] 0.2185 [1.3694] -1.8245 [-0.8400] -0.8305 [-0.1411]ΔLSENSEX(-2) 0.2792 [1.6630] 0.0491 [0.3120] -1.3976 [-0.6518] -0.6749 [-0.1161]ΔLSENSEX(-3) 0.2638 [1.5378] 0.1277 [0.7932] -1.2341 [-0.5632] -2.5071 [-0.4222]ΔLSENSEX(-4) 0.3550* [2.0592] -0.0007 [-0.0479] -0.4866 [-0.2209] 1.5020 [0.2516]ΔLSENSEX(-5) 0.4342** [2.4562] 0.3360* [2.0258] -0.9397 [-0.4162] 1.8215 [0.2977]ΔLSENSEX(-6) 0.4896** [2.7097] 0.4977** [2.9356] -3.4863 [-1.5106] 2.6972 [0.4312]ΔLSENSEX(-7) 0.3110 [1.7642] 0.1751 [1.0587] 2.2621 [1.0046] 7.3674 [1.2073]ΔLNIFTY(-1) 0.1379 [0.7003] 0.1187 [0.6426] 0.1072 [0.0426] -1.6534 [-0.2425]ΔLNIFTY(-2) 0.1044 [0.5615] 0.1172 [0.6717] 0.3358 [0.1413] 3.0635 [0.4759]ΔLNIFTY(-3) -0.2266 [-1.2327] -0.1577 [-0.9143] 2.2856 [0.9735] -3.4846 [-0.5476]ΔLNIFTY(-4) -0.3920* [-2.1079] -0.0135 [-0.0777] -0.1960 [-0.0825] -4.6127 [-0.7167]ΔLNIFTY(-5) -0.4208* [-2.1751] -0.2946 [-1.6231] 1.2228 [0.4949] 2.1918 [0.3273]ΔLNIFTY(-6) -0.2739 [-1.3844] -0.2665 [-1.4356] 5.8095* [2.2993] 3.7726 [0.5509]ΔLNIFTY(-7) -0.2160 [-1.1465] -0.3235 [-1.8296] -0.8278 [-0.3440] -1.5066 [-0.2310]ΔGSEC-10(-1) 0.0026 [0.2260] 0.0041 [0.3840] 0.0696 [0.4701] 0.4240 [1.0558]ΔGSEC-10(-2) 0.0061 [0.5558] 0.0068 [0.6581] 0.0885 [0.6294] 0.1710 [0.4485]ΔGSEC-10(-3) -0.0060 [-0.5528] -0.0028 [-0.2769] -0.1596 [-1.1339] -1.0195**[-2.6728]ΔGSEC-10(-4) -0.0131 [-1.1840] -0.0012 [-0.1228] 0.2389 [1.6800] 0.0217 [0.0564]ΔGSEC-10(-5) -0.0360** [-3.2235] -0.0277** [-2.6443] 0.2175 [1.5233] 0.2534 [0.6548]ΔGSEC-10(-6) -0.0068 [-0.5831] 0.0056 [0.5135] -0.1641 [-1.0889] -0.6826 [-1.6708]ΔGSEC-10(-7) 0.0092 [0.8302] -0.0015 [-0.1533] -0.0242 [-0.1705] 0.0877 [0.2281]ΔTB 15-91(-1) 0.0048 [1.1091] 0.0025 [0.6247] -0.0295 [-0.5333] -0.5184** [-3.4486]ΔTB 15-91(-2) -0.0014 [-0.3020] -0.0016 [-0.3698] -0.0452 [-0.7550] -0.2091 [-1.2886]ΔTB 15-91(-3) 0.0092** [2.7107] 0.0085* [1.9772] 0.0331 [0.5636] -0.0244 [-0.1531]ΔTB 15-91(-4) 0.0093** [2.8295] 0.0032 [0.7556] 0.0334[0.5706] 0.0900 [0.5674]ΔTB 15-91(-5) 0.0083* [2.1484] 0.0050 [1.2513] -0.0638 [-1.1728] -0.1507 [-1.0216]ΔTB 15-91(-6) 0.0048 [1.1612] 0.0024 [0.6186] 0.0688 [1.2791] 0.0325 [0.2227]ΔTB 15-91(-7) -0.0041 [-1.1211] -0.0024 [-0.7162] 0.0318 [0.6789] 0.1976 [1.5561]The numbers in parentheses denote lags of variables and the numbers in [ ] are t-ratios. ** and * imply significance at 1 per cent and 5 per cent levels, respectively. L denotes the logarithmic of the variable andΔ indicates the first difference of the variable.Table 3: Johansen Tests for Cointegrating RelationshipsNull Alternative Test 5 % 1 % Statistics Critical Values Critical Valuesλtrace Testr = 0 r > 0 69.97* 62.99 70.05r ≤ 1 r > 1 36.77 42.44 48.45r ≤ 2 r > 2 18.70 25.32 30.45r ≤ 3 r > 3 6.78 12.25 16.26λmax Testr = 0 r = 1 33.19* 31.46 36.65r = 1 r = 2 18.07 25.54 30.34r = 2 r = 3 11.92 18.96 23.65r = 3 r = 4 6.78 12.25 16.26r indicates the number of cointegrating relationships. * indicates rejection of the null hypothesis of no cointegration at the 5 per cent critical level. 
